A two-stage procedure for bioequivalence studies.

A two-stage experimental procedure for bioequivalence studies is proposed. The procedure is based on the idea that information from a first-stage experiment can be used to form a predictive distribution for the outcome of a second-stage experiment, thus permitting an assessment of the probability of a successful overall outcome. A systematic numerical study of a variety of possible strategies results in the identification of procedures that lead to substantial increases in efficiency compared with single-stage studies.